1.2 How CSP Works

To help you understand how CSP works, consider the following scenario of an ASR auto-attendant application.

A caller telephones XYZ company, listens to the welcome greeting (the outgoing prompt), and interrupts the prompt by speaking the name of the person she wishes to contact, for example, "Steve Smith" (this is the incoming signal).

If the CSP software detects energy above the configurable threshold in the incoming signal, CSP then terminates the prompt (this is barge-in), removes echo from the incoming signal and forwards the "Steve Smith" signal, with the pre-threshold speech that has been buffered, to the ASR application. The application recognizes the name, correctly responds to the request, and connects the caller to the intended audience.

Note:

The CSP software does not include an ASR or text-to-speech (TTS) resource.

Several CSP components (listed below) make this connection possible. Many of these components reside in the firmware level of the CSP-capable board.

• Echo Canceller
• Voice Activity Detector (VAD)
• Pre-Speech Buffer
• Barge-In and Voice Event Signaling
• Recording or Streaming to the Host

Figure1-1 depicts the data flow from the network to the CSP voice channel. It also shows how echo is introduced in the signal in the network and how it is cancelled. On DM3 boards, the option of sending the echo-cancelled signal over the TDM bus to another board is not available.

Additional functionality is provided that allows you to send a reference signal from another device over the TDM bus to the CSP voice channel. This external reference signal can be from a secondary network device. Figure1-2 depicts this feature.

Figure 1-1. CSP Components and Data Flow

Figure 1-2. Echo Cancellation Using an External Reference Signal

In Figure1-2, caller A and caller B are in conversation. There is a full-duplex connection between these two callers. Caller A's signal is received on the same physical device on which the echo canceller is located. Caller B's signal is from a secondary network device.

The CSP voice channel listens to two signals: the incoming signal from caller A (which contains echo) and the incoming signal from caller B (which serves as an external reference signal to reduce the echo).

By using caller B's incoming signal as a reference signal, the echo canceller produces an echo-cancelled signal for caller A.

Click here to contact Dialogic Customer Engineering